What is a Rollback?
Rollback, told as a story about not fixing things on the spot — just going straight back to yesterday's version.
Something's off, and fixing it on the spot makes it worse.Tinkering with it before you know what's wrong only tangles it further.
Rushing to fix it on the spot only invites more mistakes.The more you rush, the clumsier it gets, and the longer it takes.
Don't fix it in place — revert first.Revert first, then calmly find out why.
Today's version (broken) → revert → yesterday's version (fine).The code reverts like this — all at once.
Code can revert, but data needs care.The guest-log gap that built up in between sometimes can't be reverted.
Rollback = when something's off, don't fix it in place — revert to yesterday's version first, then calmly find out why.
A response strategy where, if something breaks right after a deploy, you revert to the last stable version first rather than debugging on the spot. It minimizes recovery time, leaving root-cause analysis for a calmer moment afterward.
When grown-ups say it
- Rollback
- Going straight back to yesterday's version. Recovery comes first — root-cause analysis comes later.
- Release versioning
- Keeping yesterday's version safely stored. Good storage is what makes reverting easy.
- Roll-forward
- Fixing forward instead of reverting back. Slower than rollback, but sometimes the only option.
- Easy-to-revert design
- Building things so you can always go back to yesterday. Prepared ahead of time, so reverting is fast when it matters.
- vs. canary
- Testing on 1%, versus reverting everything. A bad canary leads straight into a rollback
- vs. blue-green
- Flipping the sign back sends you right back. Blue-green makes rollback especially fast
- Data migration risk
- Records that don't follow the code back. The guest log doesn't revert as simply as the code does.
- Checkpoint
- A marked point to revert back to. The more often you mark one, the less you lose.